accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From William Slacum <wilhelm.von.cl...@accumulo.net>
Subject Re: Chain Jobs and Accumulo.
Date Mon, 16 Jul 2012 22:21:04 GMT
You'd basically be doing a copy of the getInputSplits and getRecordReader
methods, except they'd be returning the mapred version of those classes.

On Mon, Jul 16, 2012 at 3:13 PM, Juan Moreno
<jwellington.moreno@gmail.com>wrote:

> How hard would it be to implement own version using the mapred API ?
>
> Would I have to do something as complex as InputFormatBase ? (It's a
> mammoth class)
> On Jul 16, 2012 5:53 PM, "William Slacum" <wilhelm.von.cloud@accumulo.net>
> wrote:
>
>> mapred was deprecated as of 0.20.0 (
>> http://hadoop.apache.org/common/docs/r0.20.0/api/org/apache/hadoop/mapred/InputFormat.html)
>> :)
>>
>> On Mon, Jul 16, 2012 at 2:49 PM, Juan Moreno <
>> jwellington.moreno@gmail.com> wrote:
>>
>>> The hadoop API is very confusing in that regard. Currently Accumulo runs
>>> atop 0.20 and in that version , mapred is the new one and mapreduce is the
>>> old one. InputFormat currently makes use of mapreduce.
>>>  On Jul 16, 2012 5:40 PM, "Billie J Rinaldi" <billie.j.rinaldi@ugov.gov>
>>> wrote:
>>>
>>>> On Monday, July 16, 2012 5:27:09 PM, "Ed Kohlwey" <ekohlwey@gmail.com>
>>>> wrote:
>>>> > I would suggest spending the effort porting chainmapper to the new API
>>>> > (mapreduce) since the old API will eventually be removed.
>>>>
>>>> I assumed that would be true since the old API was deprecated, and that
>>>> is why we no longer support it.  However, the old API has been undeprecated
>>>> since 0.20.205.0 and 1.0.0, which seems to indicate it's not going away.
>>>>  Does anyone know what the plan for it is?
>>>>
>>>> Billie
>>>>
>>>>
>>>> > Sent from my smartphone. Please excuse any typos or shorthand.
>>>> > On Jul 16, 2012 5:22 PM, "Billie J Rinaldi" <
>>>> > billie.j.rinaldi@ugov.gov > wrote:
>>>> >
>>>> >
>>>> > On Monday, July 16, 2012 5:02:52 PM, "Juan Moreno" <
>>>> > jwellington.moreno@gmail.com > wrote:
>>>> > > Hi there, I have a use case where I need to use a Chain Mapper
>>>> > > and/or
>>>> > > Reducer. The problem is that
>>>> > > the AccumuloInputFormat extends hadoop.mapreduce.InputFormat rather
>>>> > > than implementing hadoop.mapred.InputFormat
>>>> > >
>>>> > >
>>>> > > Trying to make use of org.apache.hadoop.mapred.lib.ChainMapper
>>>> > > does not work because it requires the use of the mapred package.
Is
>>>> > > there a version of the AccumuloInputFormat which uses
>>>> > > the hadoop.mapred package instead? Can InputFormatBase be rewritten
>>>> > > with the newer API?
>>>> > >
>>>> > >
>>>> > > Thanks!
>>>> > > Juan
>>>> >
>>>> > I opened ACCUMULO-695 to add support for the old mapred API.
>>>> >
>>>> > Billie
>>>>
>>>
>>

Mime
View raw message